Attenuating Gain into LC Network- How to do it?

Hi,

I am feeding a network of LC filters (input impedance of 600R) from an op-amp, wired in an inverting config as a summing amp. There are 50k pots, feeding 47k summing resistors - into an op-amp with a 47k FB resistor.

The problem is that the LC network saturates easily, creating a nasty odd harmonic spike (I've proven this in isolation, so it's not the preceding op-amp). Therefore, I need to attenuate the level by around 10dB. I have experimented with an FFT and know this works perfectly, with a decent overall THD figure, albeit a compromise on SNR.

So...what do you think is the best way to lose this 10dB? I've tried putting a 600R L-pad in front of the LC network and it messes with the FR - creating a spike in the 250-ish region, and HF roll-off (about 1.5dB by 20k).

An easy way is to simply increase feedback. However, this will change the overall IP impedance of the summing amp, and I'm feeding it with 50k pots which can't be changed.

TIA

Justin

> best way to lose this 10dB?

Change 47K to 14.9K.

>  simply increase feedback. However, this will change the overall IP impedance of the summing amp

"an op-amp, wired in an inverting config as a summing amp. There are 50k pots, feeding 47k summing resistors - into an op-amp with a 47k FB resistor."

As described, the summing impedance is so very close to zero that you can't see it. Say a 5532 chip, gain over 1,000 below 5KHz, the feedback resistor looks like about 47 ohms. Compared to your summing resistors, this is "zero". Making it 15K resistor, 15 ohms at summing node, is zero-er.

The only reason to make your feedback resistor similar to your summing resistors is because they are usually similar gizmos and you usually want roughly the same level out of the summer as out of the channels. Give or take: multiple inputs add-up, but faders drop-down.... "unity" is often a good simple design.

But here your post-summer gizmo is some too-teeny coil that goes blattt at op-amp levels. So you may as well lose 10dB in the summer, then find 10dB gain after the L-C network. Assuming your L-C is noiseless and you have more than a few channels into the summer, there isn't even a noise penalty.

Thanks, PRR - makes perfect sense.

If I stick a transformer on the O/P, wired to step up a few dB (assuming my buffer's happy to drive it), then noise penalty should be minimal. The biggest concern will be ensuring the enclosure screens the LC network so it doesn't pick up LF noise.

J